WO2016071659A1 - Analyse de messages en mode texte échangés entre des patients et des thérapeutes - Google Patents

Analyse de messages en mode texte échangés entre des patients et des thérapeutes Download PDF

Info

Publication number
WO2016071659A1
WO2016071659A1 PCT/GB2014/053311 GB2014053311W WO2016071659A1 WO 2016071659 A1 WO2016071659 A1 WO 2016071659A1 GB 2014053311 W GB2014053311 W GB 2014053311W WO 2016071659 A1 WO2016071659 A1 WO 2016071659A1
Authority
WO
WIPO (PCT)
Prior art keywords
text
patient
topic
words
characteristic
Prior art date
Application number
PCT/GB2014/053311
Other languages
English (en)
Inventor
Guy James PROCTOR BEAUCHAMP
Ann Gail Hayes
Christine HOWES
Rosemarie MCCABE
Barnaby Adam PERKS
Matthew Richard John PURVER
Sarah Elisabeth BATEUP
Original Assignee
Ieso Digital Health Limited
Queen Mary University Of London
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ieso Digital Health Limited, Queen Mary University Of London filed Critical Ieso Digital Health Limited
Priority to US15/524,756 priority Critical patent/US11031133B2/en
Priority to PCT/GB2014/053311 priority patent/WO2016071659A1/fr
Publication of WO2016071659A1 publication Critical patent/WO2016071659A1/fr
Priority to US17/313,332 priority patent/US11948686B2/en
Priority to US18/441,358 priority patent/US20240186006A1/en

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/09—Supervised learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present invention relates, amongst other things, to a method of analysing text-based messages sent between patients and therapists.
  • Computer-based systems for providing psychological therapy are being developed in which patients and therapists can communicate using text-based messages.
  • a (computer- implemented) method comprising:
  • the method can provide an effective and efficient way of determining characteristics of patients and/or therapists by analysing the text-based messages sent therebetween. This can enable, for example, alerting of particular situations or scenarios of concern.
  • Optional features are specified in the dependent claims.
  • Figure 1 illustrates a computer-based system for providing psychological therapy
  • Figure 2 illustrates a server included in the system of Figure 1;
  • Figure 3 illustrates a method that can be performed by the server of Figure 2.
  • the system 1 includes a plurality of computing devices 2 connectable, via one or more networks 3, to a server 4.
  • the computing devices 2 may be of any type.
  • the computing devices 2 are preferably configured to run a web browser software application.
  • Users of the computing devices 2 include patients and therapists providing psychological therapy, e.g. cognitive behavioural therapy.
  • Users of the computing device 2 may also include supervisors of the therapists.
  • the network system 3 preferably includes the Internet.
  • the server 4 preferably includes one or more processors 11, volatile and non-volatile memory 12, 13, and one or more network interfaces 14, interconnected by a bus 15.
  • the server 4 may include several units as illustrated in Figure 2 interconnected via a network.
  • the non-volatile memory 13 stores computer-readable instructions 16. When executed, the computer-readable instructions cause the server 4 to perform the functions described below.
  • the server 4 is configured to enable text-based messages to be sent between patients and therapists. At least some of the messages are preferably sent via an instant messaging system. This may be achieved in any suitable way.
  • the server 4 may provide a web interface to enable users to login and send messages.
  • the server 4 is configured to analyse text obtained from these messages.
  • the server 4 may comprise a specially configured module configured to perform this function.
  • the server 4 may be configured to taken actions, e.g. provide alerts to therapists and/or supervisors, after analysing the text.
  • the server 4 obtains text from text-based messages sent between a patient and a therapist.
  • the text is preferably obtained from messages sent by the patient and by the therapist.
  • the text is preferably obtained from messages corresponding to one session of therapy, e.g. a period during which instant messages are exchanged.
  • the text may be obtained from messages corresponding to more or less than one session or from messages send at times other than during sessions.
  • the text may be obtained from messages sent by more than one patient and/or more than one therapist.
  • the method preferably starts automatically, e.g. after detecting that a session of therapy has been completed.
  • the server 4 prepares the text obtained at the first step SlOl. This preferably involves replacing words (or sequences of words) with alternatives, wherein each alternative can replace several different words (or sequences of words). For example, various misspellings or abbreviations can be replaced by corrected/full words.
  • the step also preferably involves removing stop words, e.g. common words which do not contribute to the content such as 'the' and 'to'.
  • the server 4 determines one or more features of the text obtained at the first step SlOl and optionally prepared at the second step S102.
  • the server 4 preferably determines several features of the text.
  • the features determined at the third step S103 may include one or more values describing a level to which the text relates to a topic. There are preferably several values, each of which describes (parameterises) a level to which the text relates to a different topic.
  • the topics are preferably determined using a topic model and text obtained from other messages between patients and therapists.
  • the features may include one or more values describing an emotional state. For example, there may be a value describing positive/negative sentiment and a value describing anger.
  • the values may describe a level or a variability of the emotional state.
  • the values are preferably determined using a model obtained using supervised machine learning.
  • the model is preferably obtained using training data comprising text from another source. However, this need not be the case.
  • the features may include one or more values describing or relating to a frequency of a word or sequence of words in the text. There are preferably several values, each of which describes the frequency of a different word/sequence of words.
  • the words/sequences of words preferably correspond to a set of frequently used
  • the server 4 may determine other type of features, e.g. features relating to level of repetition, reformulation or correction, complexity of syntax or vocabulary, level of similarity between therapist and patient, (sequences of) part-of-speech tags, (sub-parts of) syntactic structures, (sequences of) dialogue act tags or other indicators of pragmatic function, etc.
  • features relating to level of repetition, reformulation or correction, complexity of syntax or vocabulary, level of similarity between therapist and patient, (sequences of) part-of-speech tags, (sub-parts of) syntactic structures, (sequences of) dialogue act tags or other indicators of pragmatic function, etc.
  • the features may be stored for later use.
  • the server 4 obtains one or more, and preferably several, further features.
  • the further features preferably include features ('previous features') of text obtained from messages sent during one or more previous sessions of the patient.
  • the previous features are preferably obtained from data stored at the server 4.
  • the further features may include features that are a function of one or more other features.
  • a feature may correspond to a difference between a feature ('a current feature') obtained at the third step S103 and a previous feature.
  • the server 4 obtains data relating to the patient (e.g.
  • the therapist e.g. an identity thereof
  • the communications therebetween e.g. session number, number of words in the text, etc.
  • the server 4 is preferably configured to provide a web interface to enable a patient to complete one or more questionnaires and to determine a score therefrom.
  • questionnaires may include, for example, questionnaires relating to depression and anxiety. This is preferably performed for each session. However, this need not be the case. For example, the analysis described herein may render this unnecessary.
  • the server 4 determines a characteristic of the patient and/or the therapist using the features obtained at the third step S103, the further features optionally obtained at the fourth step S104 and the data optionally obtained at the fifth step S105. In some examples, this is performed using a model obtained using supervised machine learning. As will be explained in more detail below, the model is preferably obtained using training data that includes text obtained from other messages between patients and therapists, and data relating to the characteristic.
  • the characteristic determined at the sixth step S106 may relate to a level of a psychological condition of the patient, a change in a level of a psychological condition of the patient, and/or a predicted level or change in level of a psychological condition of the patient at the end of therapy.
  • the psychological condition may correspond to depression and/or anxiety.
  • the level of the psychological condition of the patient may be determined based on questionnaire scores (e.g. PHQ-9 scores, GAD-7 scores).
  • the characteristic may relate to a likelihood of the patient engaging in risky behaviour, e.g. self-harm.
  • the characteristic may relate to a likelihood of the patient (not engaging with and/or) dropping out of the therapy.
  • the characteristic can take one of two values or classifications.
  • One of the two values or classifications preferably corresponds to a situation or scenario of concern, e.g. a patient not being predicted to recover or improve.
  • the characteristic can take one of three or more values or be a numerical value, etc.
  • the server 4 may determine several characteristics and may use several models. In some examples, the server 4 determines one or more characteristics from the features obtained at the third step S103 without using a model obtained using supervised machine learning. This is suitable where characteristics can be directly determined from features. In particular, as explained above, at the third step S103, the server 4 may determine one or more values ('topic values') describing a level to which the text relates to a topic. At the sixth step S106, the server 4 may determine one or more characteristics that are functions of the one or more topic values. For example, a characteristic may take one of two or more values in dependence upon a topic value being above or below one or more thresholds. The thresholds may be predetermined in any suitable way.
  • topics may relate to risky behaviour by the patient, and topic values above particular thresholds may correspond to patients being classified as at risk.
  • Topics may relate to aspects of a particular psychological therapy approach, e.g. a cognitive behavioural therapy model.
  • a characteristic may relate to a level to which a therapist follows the approach.
  • Topic values below particular thresholds may correspond to therapists being classified as not following the approach sufficiently closely.
  • the model or models used may depend upon e.g. a characteristic of the patient and/or therapist (e.g. language used).
  • the server 4 takes an action. This may involve providing an alert to a therapist and/or supervisor in dependence upon the characteristic(s) determined at the sixth step S106. For example, an alert can be provided if the characteristic is determined to have a value that corresponds to a situation or scenario of concern. Alerts can be provided in any suitable way, e.g. by way of a message or a web interface provided by the server 4 to the therapist or supervisor. Alternatively or additionally, the server 4 may store the characteristics for later use.
  • the server 4 updates the one or more models used at the sixth step S106.
  • This data may correspond to one or more scores determined from one or more questionnaires completed by the patient, as explained above in relation to the fifth step S105.
  • the data relating to the characteristic e.g. the one or more scores, are used, together with the text obtained at the first step S101 and optionally prepared at the second step S102, to make up training data to update the model.
  • the one or more models need not be updated in this way.
  • a model may be updated periodically or in response to events other than sessions, e.g. a patient dropping out of therapy.
  • a model may be updated by a user, e.g. a supervisor.
  • the server 4 may provide a web interface to enable a user to obtain an initial model or to update a model using data stored at the server 4 (messages between patients and therapists, data relating to a characteristic) as training data.
  • the topic model and/or set of frequently used words used in the third step S103 may also be updated in a similar manner.
  • the data used in the first example consisted of the transcripts from 882 Cognitive Behavioural Therapy (CBT) treatment dialogues (sessions) between patients with depression and/or anxiety and their therapists using an online text-based chat system.
  • CBT Cognitive Behavioural Therapy
  • the transcripts are from online CBT provided by Psychology Online, who deliver 'live' therapy from a qualified psychologist accessed via the internet (http://www.psychologyonline.co.uk).
  • 837 are between therapists and patients who were in an ongoing treatment program or had completed their treatment by the time the sample was collected. There are 167 patients in this sample (125 females and 42 males), with 35 different therapists (for 2 patients the identity of the therapist is unknown).
  • the number of transcripts per patient ranges from 1 to 14, with a mean of 5.0 (standard deviation (s.d.) 2.7).
  • s.d. standard deviation
  • PHQ-9 Patient Health Questionnaire
  • the PHQ-9 is the depression module, which scores each of the 9 DSM-IV criteria as '0' (not at all) to '3' (nearly every day). A higher score indicates higher levels of depression, with scores ranging from 0 to 27. It has been validated for use (see A. Martin et al. 2006. Validity of the brief patient health questionnaire mood scale (PHQ-9) in the general population. General hospital psychiatry, 28(l):71-77).
  • GAD-7 Generalised Anxiety Disorder scale
  • GAD-7 see . L. Spitzer et al. 2006. A brief measure for assessing generalized anxiety disorder: the GAD-7. Archives of internal medicine, 166(10):1092-1097) is a brief self- report scale of generalised anxiety disorder. This is a 7-item scale which scores each of the items as '0' (not at all) to '3' (nearly every day). A higher score indicates higher levels of anxiety.
  • PHQ now the PHQ-9 score of the patient for the questionnaire completed immediately prior to the consultation
  • PHQ start-now the difference between the PHQ-9 score prior to any treatment and PHQ now, i.e. a measure of progress (how much better or worse the patient is since the start of their treatment).
  • a measure of progress how much better or worse the patient is since the start of their treatment.
  • the transcripts from the 882 treatment consultations were analysed using an unsupervised probabilistic topic model, using MALLET (see A. K. McCallum. 2002. MALLET: A machine learning for language toolkit, http://mallet.cs.umass.edu.) to apply standard Latent Dirichlet Allocation (see D. Blei et al. 2003. Latent Dirichlet allocation. Journal of Machine Learning Research, 3:993-1022), with the notion of document corresponding to a single consultation session, represented as the sequence of words typed by any speaker. Stop words (common words which do not contribute to the content, e.g. 'the', 'to') were removed as usual, but the word list had to be augmented for text chat conventions and spellings (e.g.
  • Table 1 Feature sets for classification experiments
  • High level (H/L) Client gender; client age group; session number; client/agent number of words and turns used;
  • Topic Probability distribution of topics per transcript (one value per topic per transcript)
  • Word Unigrams for all words that appeared in at least 20 of the transcripts, regardless of speaker; the features were the normalised counts of each word
  • Each outcome indicator was tested with different feature sets using 10-fold cross- validation (The data are partitioned into 10 equal subsamples, and use each subsample as the test data for a model trained on the remaining 90%. This is repeated for each subsample (the 10 folds), and the test predictions collated to give the overall results. This partitioning is done by transcript: different transcripts from the same patient may therefore appear in training and test data within the same fold; the use of low-dimensional topic/sentiment features should minimise over-fitting).
  • Topics 2, 6, 9, 10, 16 and 17 are negatively correlated with PHQ scores, i.e. higher levels of these topics are associated with better PHQ (see Table 4).
  • Some of these topics involve words related to assessing the patient's progress and feedback, e.g. topic 2 includes session, goals and questionnaires, and topic 17 includes good, work and positive.
  • Others relate to specific concerns of the patient, e.g. topic 6 (worry, worrying and problem) and topic 16 (anxiety, fear and sick).
  • the top twenty words assigned to each topic by LDA, and the direction of significant correlations are shown in Table 5.
  • Topic 18 0.121 Table 5: Top 20 words per topic; correlations between topic and outcome and sentiment features ⁇ '+' denotes positive correlation, '-' negative correlation).
  • topics 4, 5, 7, 8, 11 and 18 are positively correlated with PHQ scores, meaning more talk assigned to these topics is associated with worse PHQ.
  • Several of these topics relate to specific issues, such as topic 5 (sleep, bed, night) and topic 18 (eating, food, weight).
  • Some of these topics display overlap with the previous group (e.g. topics 2 and 4 both contain words reviewing progress such as session, week, next and last); this suggests that some topics (e.g. progress or particular issues) are discussed in importantly (and recognisably) different ways or contexts (possibly different emotional valences - see below), and these differences are being identified by the automatic topic modelling.
  • greater amounts of talk in topics 2, 15 and 17 are weakly associated with better progress.
  • Topics identified above are the topics identified above as involving words related to assessing progress, and feedback. Greater amounts of talk in topic 8 (checking, OCD, anxiety, rituals) is associated with worse progress.
  • -Cross-correlations between topic and sentiment features Previous work has hypothesised that automatically derived topics may differ from hand-coded topics in picking up additional factors of the communication such as valence. To explore this on a global level (i.e. at the level of the transcript, rather than at the finer-grained level of the turn) cross-correlations between sentiment and topic were examined. This initial exploration offers support for this hypothesis, as can be seen in Table 6. For example, topics 3 and 4 both contain words relating to feelings and thoughts, but topic 3 is positively correlated with sentiment, while topic 4 is negatively correlated.
  • Results of classification experiments on different feature sets are shown in Tables 7 to 9.
  • the weighted average f-score is shown, with the f-score for the class of interest shown in brackets.
  • the class of interest is patients with high (moderate to severe) PHQ-9 scores; for PHQ start-now, the class of interest is patients who are not getting better.
  • the proportion of the data in the class of interest in each case is shown in the first column in Table 7 - note that these are not exactly 50%, but reflect the actual proportions in the data.
  • Table 7 Weighted average f-scores of outcomes using different high-level feature groups (figures in brackets are the f-scores for the class of interest; i.e. PHQ now - patients with higher/more symptomatic PHQ; PHQ start-now - patients showing no change or a worsening in PHQ)
  • Table 8 Weighted average f-scores using sentiment/topic features (figures in brackets are the f-scores for the class of interest)
  • topic/sentiment (though not as well as on the symptom measures); this suggests that there are aspects of the communication that can assist in predicting patient progress, but that they are not fully captured by the topic and sentiment information as currently defined.
  • dialogue structure or style may play a role; one possibility is to look at top and/or sentiment at a finer-grained level and examine their dynamics (e.g. are positive sentiments expressed near the start or end of a consultation linked to better progress)?
  • Table 9 Weighted average f-scores using raw lexical features (words/n-grams) using LibLINEAR (figures in brackets are the f-scores for the class of interest)
  • H/L including H/L excluding H/L including H/L excluding H/L
  • Standard topic, sentiment and emotion modelling can be usefully applied to online text therapy dialogue, although care is needed choosing and applying a technique suitable for the idiosyncratic language and spelling.
  • the resulting information allows prediction of aspects of symptom severity and patient progress with reasonable degrees of accuracy, without requiring knowledge of therapist identity. However, some measures of patient progress are predicted better with fine-grained, high-dimensional lexical features, suggesting that insight into style and/or dialogue structure may be desirable, beyond simple topic or sentiment analysis.
  • the second example generally used the same methods as the first example.
  • the data used in the second example consisted of the transcripts from 2066 sessions. This data includes the data used in the first example. Of the 2066 transcripts, 1864 are between therapists and patients who were in an ongoing treatment program or had completed their treatment by the time the sample was collected. There are 500 patients in this sample (352 females, 146 males, 2 unknown), with 64 different therapists (for 2 patients the identity of the therapist is unknown). The number of transcripts per patient ranges from 1 to 15, with a mean of 5.65. Results
  • Correlation results for the second example are shown in Tables 10 to 12.
  • the topics determined in the second example are mostly coherent word lists, and can be manually qualitatively assessed and labelled. Some of the topics are similar to topics determined in the first example. The numbering of topics in the two examples is unrelated. Some topics are correlated with severity (PHQ now) and progress (PHQ start-now) as shown Table 10.
  • Tables 13 to 16 show the results of the classification experiments in the second example.
  • Table 13 is concerned with various features that can be described as coarse-grained features. Similarly to the first example, there is some classification accuracy for severity (PHQ now), but less for progress (PHQ start-now).
  • Table 13 Weighted average f -scores of outcomes using various features (figures in brackets are the f-scores for the class of interest). Not including the H/L features leads to a 1-3% reduction.
  • Table 14 is particularly concerned with the word/n-gram features, which can be described as fine-grained features. In contrast to the coarse-grained features, the fine-grained features allow prediction of progress and also final severity (PHQ final). Table 14: Weighted average ⁇ -scores of outcomes using various features (figures in brackets are the f -scores for the class of interest).
  • Table 15 is concerned with the prediction of final outcomes, i.e. whether a patient is in- or out-of-caseness at the end of a course of treatment.
  • the features used in the model are H/L, Sentiment and Topic for the first treatment session; H/L, Sentiment and Topic for the final treatment session; and the change in Sentiment and Topic between the first and final treatment sessions.
  • the features used in the model are PHQ scores obtained at an assessment session and the subsequent first treatment session; and mean anger, topic 14 (other healthcare professionals, crises, risk, interventions) and topic 16 (positive progress, improvements) for the final treatment session.
  • Final in-caseness can be predicted with greater than 70% accuracy for features I and greater than 75% accuracy for features II.
  • the table also shows results for patients who were also in-caseness at the start of the course of treatment.
  • Final in-caseness should also be sufficiently accurately predictable using features for one or more earlier treatment sessions rather than the final treatment session.
  • Table 15 Weighted average f-scores of final outcomes using various features (figures in brackets are the f-scores for the class of interest - patients who are in-caseness at the end of a course of therapy).
  • Table 16 is concerned with predicting non-engagement and drop-out, i.e. patients not entering or staying in therapy. This applied to 148 of the 500 patients. Results were obtained using text from the assessment session only, the first treatment session only and both sessions. For the assessment session only, the features used in the model were H/L, Sentiment and Topic. For the first treatment session only and both sessions, the features used in the model were H/L, Sentiment, Topic and Words. Dropout/non-engagement can be predicted with greater than 70% accuracy when features of both the assessment session and the first treatment session are used.
  • Table 16 Weighted average f-scores of dropout outcomes using various features (figures in brackets are the f-scores for the class of interest).
  • the therapy may be of a type other than cognitive behavioural therapy.
  • the method may be applied to text-based messages sent in other contexts and for other purposes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Psychiatry (AREA)
  • Primary Health Care (AREA)
  • Epidemiology (AREA)
  • Databases & Information Systems (AREA)
  • Computational Linguistics (AREA)
  • Hospice & Palliative Care (AREA)
  • Educational Technology (AREA)
  • Developmental Disabilities (AREA)
  • Veterinary Medicine (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Child & Adolescent Psychology (AREA)
  • Psychology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Surgery (AREA)
  • Social Psychology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Treatment And Welfare Office Work (AREA)

Abstract

La présente invention concerne l'analyse de messages en mode texte échangés entre des patients et des thérapeutes. Un procédé informatisé comprend les étapes consistant à : obtenir un texte à partir de messages en mode texte échangés entre un patient et un thérapeute réalisant une psychothérapie ; déterminer au moins un élément du texte ; et déterminer une caractéristique du patient et/ou du thérapeute à l'aide dudit au moins un élément.
PCT/GB2014/053311 2014-11-06 2014-11-06 Analyse de messages en mode texte échangés entre des patients et des thérapeutes WO2016071659A1 (fr)

Priority Applications (4)

Application Number Priority Date Filing Date Title
US15/524,756 US11031133B2 (en) 2014-11-06 2014-11-06 Analysing text-based messages sent between patients and therapists
PCT/GB2014/053311 WO2016071659A1 (fr) 2014-11-06 2014-11-06 Analyse de messages en mode texte échangés entre des patients et des thérapeutes
US17/313,332 US11948686B2 (en) 2014-11-06 2021-05-06 Analysing text-based messages sent between patients and therapists
US18/441,358 US20240186006A1 (en) 2014-11-06 2024-02-14 Analysing text-based messages sent between patients and therapists

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/GB2014/053311 WO2016071659A1 (fr) 2014-11-06 2014-11-06 Analyse de messages en mode texte échangés entre des patients et des thérapeutes

Related Child Applications (2)

Application Number Title Priority Date Filing Date
US15/524,756 A-371-Of-International US11031133B2 (en) 2014-11-06 2014-11-06 Analysing text-based messages sent between patients and therapists
US17/313,332 Continuation US11948686B2 (en) 2014-11-06 2021-05-06 Analysing text-based messages sent between patients and therapists

Publications (1)

Publication Number Publication Date
WO2016071659A1 true WO2016071659A1 (fr) 2016-05-12

Family

ID=52023549

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/GB2014/053311 WO2016071659A1 (fr) 2014-11-06 2014-11-06 Analyse de messages en mode texte échangés entre des patients et des thérapeutes

Country Status (2)

Country Link
US (3) US11031133B2 (fr)
WO (1) WO2016071659A1 (fr)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019220144A1 (fr) * 2018-05-17 2019-11-21 Ieso Digital Health Limited Procédés et systèmes d'administration et de surveillance de thérapie améliorées
US10680989B2 (en) 2017-11-21 2020-06-09 International Business Machines Corporation Optimal timing of digital content
WO2021058978A1 (fr) * 2019-09-26 2021-04-01 Ieso Digital Health Limited Améliorations d'une analyse d'interactions
US11031133B2 (en) 2014-11-06 2021-06-08 leso Digital Health Limited Analysing text-based messages sent between patients and therapists
EP4064289A4 (fr) * 2021-02-02 2022-12-28 Nanjing Silicon Intelligence Technology Co., Ltd. Procédé et dispositif permettant de déterminer un schéma d'apprentissage de conseil psychologique

Families Citing this family (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9420971B2 (en) 2009-10-24 2016-08-23 Carrot Sense, Inc. Extracorporeal devices and methods for facilitating cessation of undesired behaviors
US11663409B2 (en) 2015-01-23 2023-05-30 Conversica, Inc. Systems and methods for training machine learning models using active learning
US10206572B1 (en) 2017-10-10 2019-02-19 Carrot, Inc. Systems and methods for quantification of, and prediction of smoking behavior
WO2016164484A1 (fr) 2015-04-07 2016-10-13 Carrot Sense, Inc. Systèmes et procédés de quantification et de prédiction d'un comportement tabagique
US11087880B1 (en) * 2016-04-26 2021-08-10 Express Scripts Strategic Development, Inc. Machine model generation systems and methods
US12033731B2 (en) 2016-04-26 2024-07-09 Express Scripts Strategic Development, Inc. Medical processing systems and methods
US11636368B2 (en) * 2017-10-16 2023-04-25 The Penn State Research Foundation Probabilistic multigraph modeling for improving the quality of crowdsourced affective data
WO2019113122A1 (fr) * 2017-12-04 2019-06-13 Conversica, Inc. Systèmes et procédés d'apprentissage automatique amélioré pour des conversations
CN108874768B (zh) * 2018-05-16 2019-04-16 山东科技大学 一种基于主题情感联合概率的电子商务虚假评论识别方法
KR102302071B1 (ko) * 2018-06-28 2021-09-15 연세대학교 산학협력단 우울증 발병 예측 방법 및 이를 이용한 우울증 발병 예측 디바이스
US11309082B2 (en) 2018-12-10 2022-04-19 Groop Internet Platform, Inc. System and method for monitoring engagement
US11682474B2 (en) * 2018-12-12 2023-06-20 International Business Machines Corporation Enhanced user screening for sensitive services
US11068666B2 (en) * 2019-10-11 2021-07-20 Optum Technology, Inc. Natural language processing using joint sentiment-topic modeling
CN110941953B (zh) * 2019-11-26 2023-08-01 华中师范大学 一种兼顾可解释性的网络虚假评论的自动识别方法及系统
US11257583B2 (en) * 2019-12-30 2022-02-22 Carrot, Inc. Systems and methods for assisting individuals in a behavioral-change program
US11494565B2 (en) 2020-08-03 2022-11-08 Optum Technology, Inc. Natural language processing techniques using joint sentiment-topic modeling
CN114343637B (zh) * 2021-12-29 2023-10-13 河北体育学院 基于卷积和循环神经网络运动员心理压力评估方法及系统
CA3193098A1 (fr) 2022-03-16 2023-09-16 Treasure Data, Inc. Methodes d'apprentissage automatique pour determiner une probabilite d'evenement par l'analyse de sentiments de conversations numeriques
US20240237930A1 (en) * 2023-01-18 2024-07-18 Providence St. Joseph Health Subjecting textual messages sent by patients to a machine learning model trained to predict whether they are suffering or will soon suffer from psychosis

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050228236A1 (en) * 2002-10-03 2005-10-13 The University Of Queensland Method and apparatus for assessing psychiatric or physical disorders
US20120320145A1 (en) * 2011-02-16 2012-12-20 Justin Kahn Methods and Systems for Online Counseling Sessions and Clinics

Family Cites Families (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20070166690A1 (en) * 2005-12-27 2007-07-19 Bonnie Johnson Virtual counseling practice
US20080133221A1 (en) * 2006-05-17 2008-06-05 Smith Sharon S Threat assessment based on written communication
US10032127B2 (en) * 2011-02-18 2018-07-24 Nuance Communications, Inc. Methods and apparatus for determining a clinician's intent to order an item
US8676730B2 (en) * 2011-07-11 2014-03-18 Accenture Global Services Limited Sentiment classifiers based on feature extraction
WO2014026152A2 (fr) * 2012-08-10 2014-02-13 Assurerx Health, Inc. Systèmes et procédés d'aide à la décision pharmacogénomique en psychiatrie
KR102011495B1 (ko) * 2012-11-09 2019-08-16 삼성전자 주식회사 사용자의 심리 상태 판단 장치 및 방법
US10874340B2 (en) * 2014-07-24 2020-12-29 Sackett Solutions & Innovations, LLC Real time biometric recording, information analytics and monitoring systems for behavioral health management
US20160098480A1 (en) * 2014-10-01 2016-04-07 Xerox Corporation Author moderated sentiment classification method and system
CA2966962C (fr) * 2014-11-06 2023-10-10 Ieso Digital Health Limited Systeme informatise de mise en oeuvre d'une psychotherapie
US11031133B2 (en) * 2014-11-06 2021-06-08 leso Digital Health Limited Analysing text-based messages sent between patients and therapists

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050228236A1 (en) * 2002-10-03 2005-10-13 The University Of Queensland Method and apparatus for assessing psychiatric or physical disorders
US20120320145A1 (en) * 2011-02-16 2012-12-20 Justin Kahn Methods and Systems for Online Counseling Sessions and Clinics

Non-Patent Citations (10)

* Cited by examiner, † Cited by third party
Title
A. K. MCCALLUM, MALLET: A MACHINE LEARNING FOR LANGUAGE TOOLKIT, 2002, Retrieved from the Internet <URL:http://mallet.cs.umass.edu.>
A. MARTIN ET AL.: "Validity of the brief patient health questionnaire mood scale (PHQ-9) in the general population", GENERAL HOSPITAL PSYCHIATRY, vol. 28, no. 1, 2006, pages 71 - 77, XP025038645, DOI: doi:10.1016/j.genhosppsych.2005.07.003
C.-C. CHANG; C.-J. LIN, LIBSVM: A LIBRARY FOR SUPPORT VECTOR MACHINES, 2001, Retrieved from the Internet <URL:http://www.csie.ntu.edu.tw/-cjlin/liblinear>
D. BLEI ET AL.: "Latent Dirichlet allocation", JOURNAL OF MACHINE LEARNING RESEARCH, vol. 3, 2003, pages 993 - 1022, XP002427366, DOI: doi:10.1162/jmlr.2003.3.4-5.993
H. M. WALLACH ET AL.: "Rethinking LDA: Why priors matter", NIPS, vol. 22, 2009, pages 1973 - 1981
K. KROENKE; R. L. SPITZER: "The PHQ-9: a new depression diagnostic and severity measure", PSYCHIATRANN, vol. 32, no. 9, 2002, pages 1 - 7
M. HALL ET AL.: "The WEKA data mining software: An update", SIGDKDD EXPLORATIONS, vol. 11, no. 1, 2009, pages 10 - 18, XP055058312, DOI: doi:10.1145/1656274.1656278
M. PURVER; S. BATTERSBY: "Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics (EACL)", 2012, article "Experimenting with distant supervision for emotion classification", pages: 482 - 491
R. L. SPITZER ET AL.: "A brief measure for assessing generalized anxiety disorder: the GAD-7", ARCHIVES OF INTERNAL MEDICINE, vol. 166, no. 10, 2006, pages 1092 - 1097
R. SOCHER ET AL.: "Recursive deep models for semantic compositionality over a sentiment treebank", PROCEEDINGS OF THE 2013 CONFERENCE ON EMPIRICAL METHODS IN NATURAL LANGUAGE PROCESSING, 2013, pages 1631 - 1642

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11031133B2 (en) 2014-11-06 2021-06-08 leso Digital Health Limited Analysing text-based messages sent between patients and therapists
US11948686B2 (en) 2014-11-06 2024-04-02 Ieso Digital Health Limited Analysing text-based messages sent between patients and therapists
US10680989B2 (en) 2017-11-21 2020-06-09 International Business Machines Corporation Optimal timing of digital content
WO2019220144A1 (fr) * 2018-05-17 2019-11-21 Ieso Digital Health Limited Procédés et systèmes d'administration et de surveillance de thérapie améliorées
CN112292731A (zh) * 2018-05-17 2021-01-29 怡素数字健康有限公司 用于改善治疗提供和监控的方法和系统
US12073936B2 (en) 2018-05-17 2024-08-27 Ieso Digital Health Limited Methods and systems for improved therapy delivery and monitoring
WO2021058978A1 (fr) * 2019-09-26 2021-04-01 Ieso Digital Health Limited Améliorations d'une analyse d'interactions
US20220384038A1 (en) * 2019-09-26 2022-12-01 Ieso Digital Health Limited Improvements in interaction analysis
EP4064289A4 (fr) * 2021-02-02 2022-12-28 Nanjing Silicon Intelligence Technology Co., Ltd. Procédé et dispositif permettant de déterminer un schéma d'apprentissage de conseil psychologique

Also Published As

Publication number Publication date
US11948686B2 (en) 2024-04-02
US20240186006A1 (en) 2024-06-06
US20210257100A1 (en) 2021-08-19
US11031133B2 (en) 2021-06-08
US20170323065A1 (en) 2017-11-09

Similar Documents

Publication Publication Date Title
US11948686B2 (en) Analysing text-based messages sent between patients and therapists
Howes et al. Linguistic indicators of severity and progress in online text-based therapy for depression
Corcoran et al. Using language processing and speech analysis for the identification of psychosis and other disorders
Gaut et al. Content coding of psychotherapy transcripts using labeled topic models
Kauf et al. Event knowledge in large language models: the gap between the impossible and the unlikely
Chinaei et al. Identifying and avoiding confusion in dialogue with people with Alzheimer's disease
Fellbaum et al. Towards new information resources for public health—from WordNet to MedicalWordNet
Dowlagar et al. A code-mixed task-oriented dialog dataset for medical domain
Cummins et al. TIM: a tool for gaining insights into psychotherapy
Ungless et al. Potential pitfalls with automatic sentiment analysis: The example of queerphobic bias
Denecke Sentiment Analysis in the Medical Domain
Malik et al. Differences in syntactic and semantic analysis based on machine learning algorithms in prodromal psychosis and normal adolescents
Neto Subject expression in Brazilian Portuguese: Construction and frequency effects
Workman et al. Discovering sublanguages in a large clinical corpus through unsupervised machine learning and information gain
Alotaibi et al. Review of AI-Based Mental Health Apps
Hesson et al. Breaking down barriers in pediatric mental health consultations: Understanding patients’ use of I DON’T KNOW
Jarman Combining Natural Language Processing and Statistical Text Mining: A Study of Specialized Versus Common Languages
Yelland What text mining analysis of psychotherapy records can tell us about therapy process and outcome
Denecke et al. The burgeoning of medical social-media postings and the need for improved natural language mapping tools
Benjachairat et al. Classification of suicidal ideation severity from Twitter messages using machine learning
Liu et al. Extracting knowledge from on-line forums for non-obstructive psychological counseling q&a system
Wang Automated Assessment of Psychiatric Patients Using Medical Notes
George Emergency Medical Service EMR-Driven Concept Extraction from Narrative Text
Acharya Generating Personalized Hospital-Stay Summaries for Patients
Мясоєдова Semantic and Structural Features of Translating Terms of the Professional Language of Psychology

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14811951

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 15524756

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14811951

Country of ref document: EP

Kind code of ref document: A1